ITL Prompt Batch

ComfyUI has no for-loop. It's the single most annoying thing about the engine: you can't tell it "generate these ten prompts," you queue ten runs and hope something walks the list. The classic workaround is the "Load Image Batch + increment index" trick from the batch-upscale example - but that works on a folder of images, not on text. ITL Prompt Batch is the text analog: paste a list of prompts, set ComfyUI's queue count to the number of prompts, and the node emits one positive/negative pair per run, walking the list automatically.

If you've been copying a prompt into a KSampler, generating, pasting the next one, generating - this node ends that ritual.

How it works

You build a list of prompts in a row editor (each row is a positive and negative text box side by side, in the spirit of rgthree's Power Lora Loader), and the rows serialize into a hidden prompts_json field. The format is a JSON array:

[
  { "positive": "a red fox in the snow, highly detailed, 8k", "negative": "blurry, watermark" },
  { "positive": "a neon-lit city street at night, cinematic" }
]

positive is required and must be non-empty; negative defaults to "" when omitted. The importer also accepts convenience forms - a bare string counts as positive-only, prompt works as an alias for positive, and the array can be wrapped as { "prompts": [...] }.

The ritual, which you'll internalize fast:

1. Build the list.
2. Click 🔍 Check for prompts - it validates the JSON, counts the prompts (N), and resets the index to 0.
3. Set ComfyUI's queue/run count to N and hit run.

The node walks 0, 1, 2… across the batch, advancing its index by 1 after each queued run - driven from JS via its own after-queued hook, so it's immune to the "Widget Value Control Mode" setting that breaks the naive increment-widget trick. If the index ever overshoots the list (more runs than prompts), it clamps to the last prompt instead of erroring.

The inputs that matter

• index - which prompt to emit this run (0-based). You normally don't touch it; "Check for prompts" resets it, the front-end advances it. But it's also an output, which is the useful part.
• reset_index_at_batch_start (default on) - zeroes the index at the start of every queued batch, so each batch starts from the first prompt. Leave it on; turning it off means a batch continues from wherever the last one stopped.
• delete_empty_prompts (default on) - drops rows with an empty positive on Check and just before a batch is queued, so blank rows never break a run or skew the count.
• json_in (optional) - wire a text/primitive node holding a JSON prompt list here, then press 📥 Read from JSON to append it to the rows. Ignored at run time - the rows you edit are what actually runs.

Outputs & wiring

Outputs: positive (STRING), negative (STRING), and index (INT). Wire positive/negative into your CLIP Text Encode nodes, and wire index into a SaveImage filename suffix - that's how each saved image records which prompt made it. Without that, a batch of 10 looks like 10 copies.

Install

ComfyUI Manager (search "ComfyUI-IntoTheLatent-Utils") or:

cd ComfyUI/custom_nodes
git clone https://github.com/Into-The-Latent/ComfyUI-IntoTheLatent-Utils

then restart. No extra dependencies.

One honest caveat: this node pairs especially well with the pack's own ITL Save Metadata (Civitai) node, which reads the Prompt Batch's current index at save time and records the exact line that ran - but standalone, it works with any SaveImage. Just remember the repo ships with issues disabled, so the "Check for prompts" button is your validation step; use it before queueing.